Abstract
Due to the vision of the semantic web, semantic extraction has become the topic of many studies. The difficulty of extracting semantic information from natural language documents is one of the major challenges in the development of the semantic web. The semantic extraction from Chinese documents is more difficult due to the lacking of formal definitions of grammars, and of Chinese parsers. Some studies had attempted to extract ontology from Chinese text documents; however, the ontology was limited to a specific domain, which means the semantics extracted were domain-limited. Tang & Song Chinese poetry (唐宋近體詩) is a special form of Chinese literature; it is written with regular metric patterns, and although it takes a regular format, it has rich semantics. With these characteristics, we believe it is feasible to extract semantic information from Chinese poetry with simple heuristics, independent of the domain of concepts. This research proposes techniques to extract the semantic structure from Chinese poetry, and to retrieve Chinese poetry based on semantic structure similarity. The reason we have chosen to work with Chinese poetry is that we want to take advantage of the regularity of the metric pattern of Chinese poetry. To extract the semantic structure from Chinese poetry, we have designed a set of parsing rules. With our parsing heuristics, we can parse each line in the Chinese poem into an ontology instance. We can also parse it to semantic structure of our own design, which can be used later in semantic structure similarity matching. In order to retrieve poems using structure similarity matching, we have designed a semantic structure similarity algorithm to compute the similarity between two semantic structures. The user can retrieve poems using Chinese query passages or with a semantic structure written in a specific format.